Comparative biodiversity along a gradient of agricultural landscapes

Acta Oecologico 19 (I) (1998) 47-60 / 0 Elsevier, Paris

Comparative biodiversity along a gradient of agricultural landscapes

FranFoise Burel’, Jacques Baudry2, Alain Butet’, Philip Didier Le Coeu?, Florence Dubs’, Nathalie Morvan B

e Clergeau3, Yannick Delettre4, , Gilles Paillat’, Sandrine Petit’,

Claudine Thenail*, Etienne Brune16, Jean-Claude Lefeuvre’

’ CNRY- UMR 6553, Labomtoire d Bvolution des Systimes Naturels et Modz@k, Uniuersitt! de Rennes 1, Campus de Beaulieu, 35042 Rennes cedex, France.

’ INCA-SAD Armmique, 65 rue de Saint Brieuc, 35042 Rennes cedex, France. ” INRA Faune Sauvage, CNRS UMR 6553, Uniuersiti de Rennes 1, Campus de Beaulieu, 35042 Rennes cedex, France.

4 CNRT- 7JMR 6553, Laboratoire d’Ecologie du sol et de Biologic des populations, Station biologique, 35380 Paimpont, France. ’ ENSAR, Laboratoire d’Ecologie et Sciences Phytosanitaires, 65 rue de Saint Brkuc, 35042 Rennes cedex, France.

’ INRA-Zoologie, 35670 Le Kheu, Franc?.

Received April 22 1997; acceptedJuly 22 1997

Abstract - The aim of this study is to compare biodivemity in contrasted landscape units within a small region. In western France agricultural intensification leads to changes in landscape structure: permanent grasslands are ploughed, fields enlarged and surrounding hedgerows removed or deteriorated, brooks are straightened and cleaned. South of Mont Saint Michel Bay, four landscape units have been identified along an intensification gradient. Several taxonomic groups (small mammals, birds, insects and plants) have been used to evaluate the characteristics of biodiversity along this gradient. The hypothesis that intensification of agricultural practices lead to changes in biodiversity has been tested. Biodiversity is measured by the species richness, Shannon’s diversity index, equitability and similarity indexes. Our results show that intensification of agriculture does not always lead to a decrease in species richness, but to several functional responses according to taxonomic groups, either no modification, or stability by replacement of species, or loss of species. For most of the studied taxonomic groups species richness does not vary greatly along the gradient. Depending on the landscape structure and farming systems this gradient is probably truncated and does not allow to show major changes in species richness. An alternative hypothesis is that used indexes are not sensitive enough to reveal changes in biodiversity. Nevertheless, similarity indexes reveal that sensitivity to changes varies, invertebrates being more likely to perceive the dynamics of the landscapes studied than vertebrates or plants. These points have to be taken into consideration when elaborating policies for sustainable agriculture or nature conservation. 0 Elsevier, Paris

biodiversity I landscape / agriculture I small mammals I birds/insects /plants

1. INTRODUCTION

Recent changes in agricultural production methods and policies have led to major transformations in land use patterns all over the planet [66]. These have impacts on biodiversity and are often viewed as a major threat for the future [59]. Following the concern of governments and international organizations the question arises as to how to maintain biodiversity in intensive agricultural landscapes. Management strategies are planned to develop sustainable agriculture and to promote environmentally friendly practices for nature conservation. In this context evaluations of biodiversity must be developed to compare different types of agriculture.

Most ecologists [42], as well as most agronomists [33] have, until recently, concentrated their investiga- tions at the field level. The investigators then assume

that field land cover and practices at the field level are the main factors controlling biodiversity. As the development of landscape ecology shows that landscape is a more appropriate level to analyse species dynamics [69], research on farming systems demonstrate that field cover and uses are highly dependent on the functioning of the farm they are part of [13]. The relationships between farming systems and landscape patterns have been mainly studied at a very broad scale [45] on rural European landscapes, focusing on the contrasting effects of the main farming systems. Within an area of a few thousand hectares where farm types are not very contrasted, understanding of the relationships is more complex [S], because the field mosaic of the different farms has to be analysed. As it is a proper scale for ecological investigation - species movement, population dynamics - we must develop a framework of the impact of agriculture on landscapes and biodiversity at these scales.

48 F. Burel et al.

It also has practical aspects for conservation biology. Farms are the units for the implementation of agri- environmental policies. Under the first phase of Euro- pean agricultural policies, changes in agricultural landscapes have been dramatic in western Europe over the last few decades [2, 161. The grain size has been enlarged, and many uncultivated elements, such as woodlots or hedgerows have been removed to facili- tate cultivation. At the same time, brooks have been straightened and cleaned. At the farm and field levels, both the use of high input rates (pesticides, fertilizers) and the increase of the disturbance regime influence habitat quality [43].

Few studies deal with the measure of biodiversity in agricultural landscapes. Diversity occurs at various levels of biological complexity, from genes to landscapes [26]. Species richness and diversity are commonly used for conservation purposes and to assess ecosystem fitness [52,57]. In agroecology most of the data on the negative effect of intensification on species richness have been obtained at field scale or even smaller plots scale [30, 551. Ecological studies, considering more ‘natural’ areas, (forests, coastal environments, etc.) show that species diversity is strongly correlated with habitat diversity and that it peaks over intermediate disturbance levels, when considering fine scale disturbance [58]. Environmental heterogeneity in space and time is important for maintaining species diversity [34] and is related to disturbance regimes [64]. High input agricultural systems are generally associated with homogeneous landscapes, in which woody elements are scarce [46]. In most cases, intensification of agriculture results in loss of natural habitat, but as fragmentation increases, patch size and isolation aggravate the effect of habitat loss. For this reason, Andren [4] concluded that the loss of bird and mammal species or the decline of their population size will be greater than expected from habitat loss alone. There might be a threshold in intensification of agriculture influencing species richness at the landscape level but we could expect more effects on population processes than on biodiversity itself for vertebrate taxa. All these results lead to the strong idea that intensification of agriculture is a threat for biodiversity.

This paper deals with the evaluation of biodiversity over a gradient of landscape structure correlated with different farming practices. Along this gradient we identified four landscape units. These units did not cover the whole gradient from forest to totally open landscapes, but were four stages of agricultural intensification. Not all the taxa have been recorded, which is unrealistic, but several groups have been selected: plants, invertebrates and vertebrates. Even if the hypothesis of a loss of biodiversity along the gradient is commonly accepted, alternative possibilities may be identified: (i) decrease in species richness, (ii) no

change in species richness or composition, (iii) no change in species richness but composition varies and (iv) increase in species richness.

2. METHODS

2.1. The study area

The study area is located in northern Brittany, south of Mont Saint Michel Bay, France (48” 36’ N, 1” 32’ W). A hedgerow network (bocage) characterizes three units, and the fourth is a polder structured by a network of dikes, some of which are planted with poplars (figure 1). The area varies from 500 to 1 000 ha. In the first three landscapes agriculture is orientated toward milk production while in the polder it has shifted, since 1970, towards intensive vegetable production [39]. We essentially based the delimitation of the units on landscape structure drawn from aerial photographs. We took into account the grain size of the field mosaic, the density of hedgerow network, and the relative abundance of grassland vs. cropland.

A survey of farming systems also concluded that these units are different in terms of agricultural practices: farm size, level of production. Inputs increase from site A to site D [ 1,631. Therefore, the differences in landscape structure are associated with other aspects of intensification.

2.2. Characterization of the four landscape units

The recognition of both the effects of spatial structure and scale dependence on ecological processes has stimulated a number of studies on landscape patterns at different scales [67]. Patterns can be described for their own properties [ 16, 20, 491 or in relation to land use changes [44] or the distribution or movement of a species [56]. In this paper we aim at a single landscape description for a wide range of species behaving at different scales. In this perspective, we adopted a simple description reflecting the feature commonly associated with ‘agriculture intensification’ in western France: ploughing of grassland, clearing of woodland, removal of hedgerows .

Comparisons of biodiversity among landscapes imply that (i) they are different and (ii) they have some kind of internal homogeneity. Global indices such as percentage cover of woodlots, grassland, crops, hedgerows, heterogeneity could show differences but do not provide information on the internal structure. We used a multiscale approach where the different landscapes are divided with grids of analytical windows of various size. We then characterized the structure of the landscape within the windows and evaluated these classes of structure in degree of relation to the different landscapes. The multiscale approach makes it

Biodiversity in agricultural landscapes 49

SITE A

SITE C Legend

woodland

grassland

SITE D

I-1 cropland

I1 roads & built UP area

J hedgerow 4-W ,--/ grassy strip 750 m

Figure 1. Changes in species composition from the most closed to the most open landscape.

possible to describe landscapes over a range of scales at which the different organisms may perceive them.

Maps of the four landscapes were digitized in a raster format, using IDRISI [27]. Five categories of land cover were used: woodland, permanent grassland, cropland, hedgerows and roads and built up areas. In the polders, because the ditches and non wooded dikes have a grassy vegetation, it was decided to assign them as grassland and rows of poplars as hedgerows. The maps were divided in square windows of 100, 500 and 750 m, using ChloC, a routine especially designed for this purpose [lo]. Within each window the number of pixels of each type was computed as well as the connections among adjacent pixels. The connections were

computed as follows: where i and j are types of land cover of successive pixels on a row or a column, the routine calculated the number of (ij) and then the proportion of each possible type of couples of pixels. The result indicates the number of pixels of each category and their spatial arrangement. To avoid bias resulting from the position of the grid of windows, several grids (4, 10 and 14, the number increasing with the size of the window) were used with a different starting point. In subsequent analyses we used the tables: windows X types of connections.

These raw figures were split into classes of frequen- cies to perform multiple correspondence analysis. The class limits for grassland, woodland, and cropland are

Vol. 19 (I) 1998

50 F. Burel et al.

O-25 %; 25-50 %, 50-75 % and 75-100 % presence of pixels in windows, for hedgerows, limits are O-5 %; 5- 10 %, lo-25 % and 25-100 %. Use of classes takes into account non linear relationships among types of land cover [32]. To enhance the amount of variance explained by the analysis the coding was made using fuzzy logic. Therefore the data table gives the grade of membership of a window in any given class of value [35]. A window may belong to two adjacent classes.

The correspondence analysis was performed on a set of windows of the different sizes (ANCORR program of ADDAD) [38]. The windows were then set into classes using classification techniques (CAHVOR program from ADDAD) run on the first three factors of the ordination. Finally, the relationships between classes of windows and landscape units were tested. Mutual information was used to measure the relationships between classes and landscapes. This measure is derived from information theory similar to a &* test, using Kullback’s criteria. The advantage over x- is that there is no requirement in terms of number of cases. The strength of the relationship is given by the redundancy, i.e. the percentage of the entropy (diversity in the Shannon sense) of the diversity of window types explained by their belonging to a given landscape unit.

2.3. Measures of biodiversity

2.3.1. Groups studied

Vertebrates, invertebrates and plants have been surveyed in this study. We posit that small mammals and birds should perceive changes in landscape structure. Birds have been proved to be good indicators of habitat structure and composition [l 1, 291. In heterogeneous landscapes, small mammals can move over long distances [36] and may survive in very small patches [41.

Three groups of insect species were selected: Empi- didiae and Chironomidae (Diptera) and Carabidae (Coleoptera). Empididae and Chironomidae Diptera both have an active phase of dispersion at the adult stage. Emergence, resting and swarming sites are dis- tinct elements of the landscapes, that are reached by passive or active flight depending on the species. These movements are controlled locally by the spatial distribution of hedgerows, brooks and adjacent fields [51]. &rabid beetles (Coleoptera Carabidae) either fly or walk. Flying ability is related to the size of the species and to their potentiality to react to disturbance. Forest carabid beetles, restricted to woody habitats, are flightless and not very mobile [ 181. Smaller species may fly easily and characterize crop land and early successional stages [25].

Flowering plants, including herbs of the bottom layer and shrub and trees of the ligneous upper layers

of hedgerows, were assessed over the three bocage landscapes. As less mobile organisms, they can be supposed to integrate dynamics of landscape changes.

2.3.2. Sampling methods

Sampling methods to count species in heterogeneous landscapes depend on the target species. For birds and small mammals, tools are available to cover the whole landscape units. Sampling strategies con- cerned only the temporal dimension. In this study, birds were counted according to ‘IPA’ method [12]. During the spring of 1993, passerine birds were surveyed in the three bocage units, IPA took into account adjacent squares (side = 250 m), in order to cover the whole area. This resulted in 110 counts on sites A, B and C. During the winter of 1992-1993 birds feeding in the agricultural area (excluding aquatic and marine birds) were counted in sites C and D: 8 sampling sites distributed regularly along 7 km long gradients ori- ented North-South and crossing the whole units, were surveyed 8 times in each site. To study the small mammal community of the four landscape units we used analysis of pellets of the Barn owl (Tyto a&a). The Barn owl is known to have a large prey spectrum including essentially small mammal species as rodents and shrews [61]. The diet composition of this raptor is strongly influenced by prey availability and this method has been found effective and already used to study the relative abundance of small mammal species at the landscape level [7, 191. The foraging area of this raptor is from 250 to 700 ha around the breeding or wintering site [48, 611. A roost was found in each landscape unit (all were abandoned buildings) and vis- ited five times during the year. A total of 1982 preys were used and 1947 preys were identified by comparing the skulls and jaws to the taxonomic keys in Chaline et al. [ 171. Frequency of occurrence (%) of the various species was then evaluated for each landscape.

When studying insect species, space and time have to be sampled as the only techniques available are traps of different kinds. For carabids (Coleoptera Car- abidae) interception traps were used and for the two families of Diptera (Empididae and Chironomidae), yellow attractive traps have been utilized. Traps were located for most of them along the uncultivated elements which host most of the species, as temporary refuge or permanent habitat in intensive agricultural landscapes [ 151.

For carabids, sixteen sets of pitfall traps were installed in hedgerows along a transect crossing each landscape unit in 1992. As from this date, traps were opened one week per month, i.e. 50 sampling periods. For the two families of Diptera 128 yellow traps (38, 52 and 38 in sites A, B, C respectively) were set in pairs on hedgerow ground along a transect from a brook to the middle of each study site from May to


July 1994 and opened three days each week. This period was chosen in accordance with known emergence phenology of Empididae.

As for plants, only the uncultivated linear elements were assessed, and two different sampling designs were used, in line with sampling effort. For woody plants, every hedgerow, defined as a linear feature har- bouring at least one shrub or tree, was studied on the whole area of sites A, B and C. The latter investigation resulted in a total set of 3 014 woody plants releves (1 534 in site A, 1 68 1 in site B and 819 in site C). In order not to bias data with recently planted garden species, which were introduced in newly created hedgerows of site C, only spontaneous or subspontaneous ligneous species were retained in subsequent analyses. Assessment of herbs of the bottom layer was restricted to 3 selected boundary networks, chosen in order to encompass a variety of both boundary structure (from grassy strip to dense hedgerow) and adjacent to boundary land use (mainly crop, grassland and road). Herb releves (25 m long) were made on both sides of each uncultivated linear feature of the selected network. The resulting data set consisted in a total number of 455 releves (137 in site A, 186 in site B and 132 in site C).

2.3.3. Data analysis

Several indexes may be used to compare species assemblages [41]. Species richness (S), Shannon’s diversity index (H’) and equitability index (E) are ways of comparing number and abundance of species. I will be high if S and E are high. These indexes do not account for differences in the species present, as assemblages differing in their species composition may have the same values. Similarity indexes were used to account for these variations. For each taxonomic group they were computed for the two most extreme landscape units along the gradient.

Sorensen’s community index [31] measures the similarity among two assemblages. It is based on species numbers alone and does not take species abundance into account:

I (Sorensen) = 2c/(a+b) where ‘a’ is the species number of assemblage A, ‘b’ the number of species in assemblage B, ‘c’ the number of species common to A and B.

The Squires overlapping index [60] measures similarity taking into account the abundance of the species in the two assemblages:

I (Squires) = Smin (xi,yi) where xi and yi are respectively the relative abundances of species i in A and B

Values of these two indexes vary from 0 to 1; 0 indicates that assemblages differ totally, and 1 that they are identical. The combined use of these two measures

makes it possible to compare the response of the different groups to the landscape gradient.

For each group, rates of stable, appearing and dis- appearing species, from one end to the other of the gradient was calculated. This provides a way of assessing the behavior of the groups according to our hypotheses of responses to the gradient.

3. RESULTS

3.1. Characterization of the four landscape units

Correspondence analysis: For the three analyses (100, 500 and 750 m) the eigen values of the first factor are respectively: 0.22; 0.23; 0.24 and the percentage of variance taken into account: 17.8; 48.9 and 62.1 %. For the second factor the percentage variance explained are: 9.7; 14.5 and 12.9, for the third: 7.2; 7.4 and 7.3. The contrast among windows increases as their size increases, but only on the first factor.

Cluster analysis: The three analyses yielded 6, 6 and 4 significant classes (the changes in the inertia at the aggregation of these classes show a higher threshold than in previous aggregations). The main characteristics of these classes are given in table Z, further details are in Appendix 1.

At the three scales, types of windows (within unit landscape types) and landscape units are significantly associated (redundancy varies from 23 to 58 to 76 %) (Appendix 2). The strength of the relationship increases as the scale becomes coarser. Therefore, the landscape units we studied are different not only at a global scale, but at all scales.

The polder unit is always more different than are the three other units. At the coarser scale unit A and unit B share many class 4 windows (cropland, grassland, hedgerows), but differ because unit A has windows where grassland is dominant (as unit C does) and unit B has Class 2 windows (cropland dominant). An internal heterogeneity thus always shows up.

3.2. Data analysis

Species richness, diversity and evenness for the different groups are given in table II. Depending on the groups, these descriptors vary in different ways when confronted with the gradient of intensification estab- lished from landscape structure analyses. A clear decrease in species richness only appears in Diptera Chironomidae and Empididae at site C. Linear rela- tions appear clearly only for Diptera Empididae and breeding passerines. In the other groups, richness remains stable (small mammals), changes very slightly (wintering birds, woody plants) or fluctuates with no relation with the gradient (carabid beetles, herbs). Variations in diversity indexes are not necessary asso-

Vol. 19 (I) 1998

52 F. Burel et al.

Table I. Main characteristics of types of windows at different scales

Class I Class 2 Class 3 Class 4 Class 5 Class 6

100 x 100 m windows

grassland and a few

hedgerows

75 % grassland,

25 % cropland and

above average

hedgerows, mostly

around grassland

500 x 500 m windows

almost exclusively

cropland, very few

hedgerows

mostly cropland with

hedgerows

750 x 750 m windows

almost exclusively

cropland, very few

hedgerows

cropland dominant, grassland dominant,

hedgerows hedgerows

a mixture ofwoodlots. almost only cropland, mostly cropland with almost only cropland

grassland and cropland no hedgerow hedgerows (more with hedgerows

frequent around

grassland)

grassland is dominant. equal share of one third grassland, mostly cropland and

hedgerows moderately grassland and two third cropland. less hedgerows

frequent cropland, hedgerows hedgerows

more frequent around

grassland

cropland dominant

many hedgerows along

grassland

Table II. Richness (R), Diversity (D) and Evenness (E) of plant and animal communities in the four landscape units

Site A Site B Site C Site D

S H’ E S H’ E S H’ E S H’ E

Diptera chironomidae

Diptera empididae

Carabid beetles

Wintering birds

Breeding birds

Small mammals

Herbs

Woody plants

28 2.56 0.53 29 I .68 0.35 I5 I.85 0.47

x4 3.69 0.58 82 3.77 0.59 56 3.27 0.56

55 2.52 0.43 51 2.50 0.44 50 2.54 0.45 58 2.67 0.45

43 3.19 0.59 58 2.67 0.45

40 4.33 0.8 I 35 4.32 0.84 32 4.32 0.86

II 2.93 0.84 II 2.7 0.80 I I 2.96 0.85 I I 2.64 0.76

I89 6.56 0.87 I32 6.06 0.86 I71 6.42 0.87

40 3.99 0.75 41 4.14 0.77 38 4.17 0.79

ciated with these trends and do not give more infor- mative explanations concerning the reaction of the different taxa to agricultural intensification. A higher evenness is observed in plant and vertebrate groups than in invertebrate ones.

Similarity indices @gure 2) provide another inter- pretation of the behavior of the different groups studied. Each group reacts specifically and its position on a gradient of similarity will depend on both quali- tative and quantitative changes affecting the specific composition of the community. Similarity indices discriminate invertebrates from vertebrates and plants, indicating that more changes occurred in the species composition or abundance of the former. Herbs and

woody plants (non spontaneous ones being excluded) are no more affected by intensification than vertebrate animals.

Although responses of the various groups are unique, they can be grouped according to the four possibilities of evolution of biodiversity yigure 3). An increase in biodiversity is only observed in the case of carabid beetles, for which the number of new species exceeds losses. The maintenance of species richness hides two ecological behaviors of the communities in response to intensification of agriculture. Breeding passerines, woody plants and small mammals can be grouped as taxa being poorly affected and sharing always a high proportion of species. Wintering birds, carabids beetles


I! High similarity 1 0.9 $

g 0.8 --

z 'GO.7 --

-50.6 --

2 r 0.5 --

$04-- 0 --

Chironomidae I Woody plan&I

Breeding passerines E Herbs I

I Small mammals I

Wintering birds I

Empididae I

0.5 0.6 0.7

Community i0nsdex 0.9 1

Figure 2. Overlapping index (Squirres) vs. Community index (Sorensen) for each plant and animal community. Data from the less to the most intensified landscape were used to compute indices.

80%

60%

lironomidae bantering birds Herbs Woody plants: - - ^ tmplaiaae trreeomg passermes

I small mammals

/ @jj Shared sp. 1 J Lost sp. M Gained sp. 1

Figure 3.750 x 750 m windows characteristic of each landscape unit.

and herbs may be considered as groups maintaining their diversity by compensating for loss of species by gain of new ones in more intensified landscape. Diptera

Chironomidae and Empididae are the two groups dis- playing a real decrease in their diversity as intensification of agriculture increases.

Vol. 19(l) 1998

54 F. Bud et al.

4. DISCUSSION

4.1. Landscape patterns

The definition of the landscape units where san- pling is carried out is a critical issue. We chose a short part of the gradient, going from forest,to treeless agricultural land. This segment of the gradient is a realistic view of the possible changes during land planning processes such as realotment programs. The visible structure of the landscape to the eye of the researcher. on the ground or from aerial pictures 131, is the only way to LL priori segregate landscape units. The definition of simple criteria to characterize these units is needed. Many authors [9,6.5] have proposed more comprehen- sive indices to describe patterns. The main problems they pose is that their value is scale dependent and the taxonomy used to describe the landscape elements should be species dependent (some species react to some types of crops, others to other types). For this reason, we utilize a classification based on multi- variate information. With simple measures combining the frequency of the different elements and their spatial relationships, we show differences at different scales. The originality of the method is to considet both patches and linear features in the description. Usually the focus is on either one. When a corridor effect is looked at, linear features are analysed [14]: when fragmentation of mosaic effects are the focus. only patches are considered [28]. In the latter case, the spatial scale of investigation is generally too coarse to allow the representation of linear features. It is important to note that in the units presented here, there is no linear relationships between abundance of a given type of patch (e.g. grassland) and linear features (hedgerows). The global differences as well as the internal structure can be evaluated. Even if the landscape units are different at all scales, the fact that, at fine scales, similarities of elements and pattern exist indicates that species requiring only small habitats may be present in all four units. The differences at coarse scale indicate that the small elements are distant; the possibilities of movement among them, either for recolonization after extinction or to achieve the different life stages of each individual (for invertebrates) will be a critical factor.

4.2. Biodiversity measurements

To evaluate the actual number of species in heterogeneous and diverse landscape units very large sam- ples are necessary as they may differ from observed numbers of species when rare species are numerous [37], so our results only permit comparison between sites where the same sampling procedures have been used. Our results show that species richness and diversity do not always respond to changes in landscape structure and quality. Depending on the taxa studied,

some follow the intensification gradient, some do not. Beside species richness, Shannon diversity index does not allow to discriminate more clearly the landscape units, for any of the studied taxa. This means that relative abundance of the species does not vary much as species numbers change. It is difficult to find any ecological meaning to the stability of the values, as this measure has a rather tenuous foundation in ecological theory 1371.

Assemblages studied vary by their species richness, from 1 I for small mammals to 189 for herbs. Evenness is lower for invertebrates than for vertebrates and plants. This may be related to some of the results on species composition in the different units, groups with high value of eveness exhibit no or few changes in species composition, while for the others there is a loss OI replacement of some species. In assemblages where there is a high proportion of rare species sensitivity to changes seem to be higher, those species being more likely to disappear.

4.3. Loss of diversity

In Chironomidae two different types of species coexist: whilst most of them inhabit water bodies at the larval stage, some species are however clearly terrestrial and their larvae are found in many different soil types [S]. In this study, the high rate of species loss from site A to site C (resulting in a low Sorensen’s community index) is mainly due to aquatic species, the number of which is divided by two. Furthermore, landscape structure interacts with species flying abilities to modify their dispersal range [23].

In contrast, terrestrial species are mainly located in wet and undisturbed soils such as permanent or long- lasting temporary grasslands. Their larval development mainly depends on the soil water regime [21] and ploughing affects negatively their larval abundance [22]. Nevertheless, the same most common terrestrial species are found in the three sites currently studied.

As a result, the high overlapping Squires’ index is related to the fact that species shared by the sites are very common and widespread. The resulting picture for the whole set of species is an obvious impoverish- ment of the chironomid community (both in species quality and quantity) in the most open and disturbed site C.

Diptera Empididae exhibit a strong species loss. The difference is less noticeable between site A and site B. At every scales, the landscape characterization clearly distinguish site C from the two other sites of bocage (A and B). Thus, the landscape structure appears as an important factor in explaining the biodiversity of Diptera Empididae. Empid species react to landscape structure on a range of fine scales and are dependent on the nature of the cultivated mosaic and on the veg-


etation structure of hedgerows [SO]. Similar results have been found with Diptera Dolichopodidae trapped in Malaise tents in the three bocage sites (Brunel, unpublished data).

4.4. Maintenance of diversity by species replacement

Numbers of carabid species do not vary much along the landscape gradient. In hedgerow network landscapes carabid species may be either restricted to one habitat (hedgerows or fields), or use both elements alternatively: field during the growing season and margins as overwintering sites. Along the intensification gradient there is a shift in the assemblage. Importance: of forest species decreases as the woody covet decreases, and hedgerows become more isolated. Car- abid beetles hibernating in margins find less favour- able conditions as field size increases. Open field species, adapted to high rates of disturbance appear. The average size of species decreases, and there is a replacement of long species by smaller ones, ubiqu- ists, more common and characterized by a shorter life cycle.

A similar trend is observed for wintering birds. Typ- ical wintering species (plovers, lapwings) of open landscapes appear in the most intensified unit. But, when looking in detail at species identity, we note that specialist species such as woodpeckers, tits, tree- creepers and nuthatches closely associated to large trees are mainly replaced by more generalist and often widespread ones.

Assemblages of hedgerow herb species show a similar reaction along the gradient of intensification. Widespread light demanding species of grassy environments are shared by site A and C. The pattern of replacement of herb species along the gradient is partly related to a shift from perennial shade demanding plants, including the so called forest herb species. to broad niche species, including opportu- nistic annuals.

4.5. Maintenance of diversity by no or few changes in species composition

For breeding passerine birds there is a slight decrease in species richness as the density of tree cover decreases from dense hedgerow network landscapes to more open ones. Even if the change is small due to the narrow landscape gradient, this is consistent with many studies on bird diversity, at various spatial scales. MacArthur [40] found that bird species diversity in forests and fields of north-eastern USA fits habitat diversity index based on vegetation layers. In agricultural landscapes Balent and Courtiade [6] found that the importance of tree cover within 2.5 ha quad- rats, was (with land use spatial heterogeneity) the best

descriptor of birds species assemblages. Decrease in habitat diversity induced by agricultural intensification may be related to decrease in species richness and diversity on a broad variety of scales.

Species richness of small mammals does not vary along the landscape gradient. The eleven species are the same in all the units, and counting them or looking at their identity does not highlight any difference in landscape functioning. As for small mammals, many data on local extinction, but rarely at the landscape scale, are reported in the literature [47]. In the case of strictly habitat-dependent species, local extinction may occur in small or isolated patches [53, 681 but species display high ability to colonize vacant territories and are generally prolific. Moreover, generalist species such as the wood mouse move through a variety of habitats depending on food availability [62]. They can use farmland mosaics and are well-adapted to agricultural landscapes, provided that hedgerow area is sufficient to spend winter (Paillat and Butet, in press) For these rea- sons species richness of the small mammal community appears as an inappropriate indicator of agriculture intensification. Only equitability and overlapping indices indicate that small mammal show differences in assemblage structure between the four landscape. This appears clearly in the fourth landscape unit where species such as the white-toothed shrew or the root vole, more adapted to open and intensified areas, have increased. Inversely species more or less dependent on woody habitats (bank vole, wood mouse) exhibit reg- ular decrease of their abundance along the gradient. As a result, evenness of the small mammal community tend to decrease with intensification of agriculture. Moreover the analyses of temporal fluctuation of the different species (unpublished data) indicate that in most cases, populations become more unstable as intensification increases. Delattre et al. [24] reported similar results when studying common vole population in eastern France. Similarly, Paillat and Butet [53] show that fluctuation range of the bank vole increases with hedgerow fragmentation. We may conclude that only severe intensification of agriculture can induce marked alterations of the diversity of small mammals such as mice vole and shrew.

Shrub and tree species assemblages exhibit very few changes in both richness and composition along the gradient of agriculture intensification. The latter trend can be related to the fact that, as hedgerow species, most of them, especially in mesic environments are likely to have been planted, or at least favoured. So, direct, past and present management of the hedgerow has more influence on species composition that agricultural practices at the landscape level. Nevertheless, rare forest tree species, as Pyrus cordata or Tilia cordata were only found in dense hedgerows of site A. If intensification of agriculture leads to the attenuation of

Vol. 19 (I 1 199X

56 F. Burel et al.

the forest atmosphere of linear woody structures, via opening of gaps or decreasing in woody layers cover, it is likely to threaten species that are sensitive to this aspect of habitat quality.

5. CONCLUSION

The results of this study force reconsideration of the impacts of agricultural activities on biodiversity. There is no simple linear relationship between intensification of agricultural land use and loss of species. Several points have to be taken into account. First, agriculture operates at several spatio-temporal scales and thus has more or less direct impact on species, depending on their life history traits. Second, agricultural landscapes are heterogeneous and disturbances due to agriculture do not induce similar rates of change for all elements. Third, agricultural landscape units have to be defined according to human activities, but to manage them in a conservation perspective will necessitate drawing other limits coherent with biological spatial units.

Another implication for conservation biology is that, for the moment, most policies such as level of chem- ical input, the stocking rate, date of haying, grazing are meant to be implemented at the field scale or to affect farm structure (cattle size, area of grassland). There is no differentiation according to the landscape context, nor plans to design the new practices within a spatial context

There is no way to undertake spatio-temporal studies of all the taxa in a broad range of landscape units. Simulation studies based on real landscapes and on functional groups of species, based on their life history traits are the tools we are currently developing to go deeper inside the comprehension of the relationships between landscape dynamics and biodiversity in agricultural areas.

Acknowledgements

The authors thank D. Vollant, V. Adamandidis and L. Lunel for their help during the field work, the identification of beetles and analysis 01 barn owl pellets. D. Denis provided help for the design and the management of the databases. Thanks to Prof. R. Spittal for English editing. This research has been financially supported by the Minis&e de I’Environnement (ComitC Ecologic et Gestion du Patrimoine Naturel), the European Commission (Programme Fair IV) and the CNRS. ComitC Systemes ruraux du Programme Environnement.

REFERENCES

[I ] Acx A.S., HCttrogCn&itC spatiale des pratiques agricoles dam les polders du Mont Saint Michel, DAA Genie de I’Environne- ment, Rennes, 1991,45 p,

[2] Agger P., Brandt J., Dynamics of small biotopes in Danish agricultural landscapes, Landscape ecol. I ( 1988) 227.240.

[3] Allen T.F.H.. Hoekstra T.W., The confusion between scale- defined levels and conventional levels of organization in ecology, J. Veg. Sci. I (1990) 5-12.

[4J Andren H.. Effects of habitat fragmentation on birds and mammals in landscapes with different proportions of suitable habitat: a review. Oikos 71 (1994) 3.55-366.

[5] Armitage P., Cranston P.S.. Pinder L.C.V., The Chironomidae. The biology and ecology of non-biting midges, Chapman & Hall. London. 1995, 572 p.

[6] Balent G.. Courtiade B., Modelling landscape changes in a rural area of south-western France, Landscape Ecol. 6 (1992) 1X3-194.

]7] Barbosa A.. Lopez-Sanchez M.J., Nieva A., The importance of geographical variation in the diet of Tvto u/ha Scopoli in central Spain, Global Ecol. Biogeogr. 2 (1992) 75-X I,

[X] Baudry J., Landscape dynamics and farming systems: problems of relating patterns and predicting ecological changes. in: Bunce R.G.H.. Ryszkowski L.. Paoletti M.G. (eds), Landscape Ecology and Agroecosystems, Lewis Publishers, Boca Raton. 1993, 2 l-40.

[Y] Baudry J., Baudry-Burel F., La mesure de la diversitt spatiale : utilisation dans les evaluations d’impact, Acta Oecol. Oecol. Appli. 3 (1982) 177-190.

] IO] Baudry J., Denis D.. Chloe: A routine for analysing spatial heterogeneity (IDRISI image files). INRA. SAD-Armorique. 19%.

] I I ] Blonde1 J., Biogeographic Cvolutive, Masson, 19X6, 22 I p. ( 121 Blonde1 J.. Ferry C., Frochot B., La methode des indices ponc-

tuels d’abondance (IPA) ou des releves d’avifaune par “station d’ecoute”. Alauda 38 (1970) 55-7 I,

] 131 Brassier J., De Bonneval L., Landaia E.. System studies in agriculture and rural development, in: Science Update, INRA Edi- tions. Paris, 1993, 415 p.

114) Burel F., Effect of landscape structure and dynamics on species diversity in hedgerow networks, Landscape Ecol. 6 (1992) 161-174.

] IS] Burel F., Hedgerows and their role in agricultural landscapes. Crit. Rev. Plant Sci. I5 (1996) 169-190.

] 161 Burel E, Baudry J., Structural dynamic of a hedgerow network landscape in Brittany France, Landscape Ecol. 4 (1990) 197-210.

[ 171 Chaline J., Baudvin H., Jammot D., Saint-Girons M.C.. Les proies des rapaces (petits mammiferes et leur environnement). Doin ed., Paris, 1974.

[IX] Charrier S., Petit S., Burel F., Movements of Ahtrx pcrmllelrpi- /r&us (Coleoptera carabidae) in woody habitats of a hedgerow network landscape: a radiotracing study. Agr. Eco. Env. 61 (1997) 133-134.

]lY] Cooke D., Nagle D., Smiddly P., Fairley M.R.I.A., 6 Muirc- heartaigh 1.. The diet of the barn owl Qfo &a in County Cork in relation to lande use, Proceedings of the Royal Irish Academy 96 ( 1996) 97-I I I.

[20] Cullinan V. I., Thomas J. M., A comparison of quantitative methods for examining landscape patterns and scale, Landscape Ecol. 7 ( 1992) 2 I I-227.

[2l] Delettre Y.R., Flux d’evaporation corporelle et resistance a la dessiccation chez les larves de quelques Chironomidae ter- restres (Diptera), Rev. Ecol. Biol. Sol 25 (1988) 129-138.

122) Delettre Y.R., Lagerlof J., Abundance and life history of terrestrial Chironomidae (Diptera) in four Swedish agricultural crop- ping systems. Pedobiologia 36 (1992) 69-78.


[23] Delettre Y.R., Trehen P., Grootaert P., Space heterogeneity, space use and short dispersal in Diptera: a case study, Land- scape Ecol. 6 (1992) 175-181.

[24] Delattre P., Giraudoux P., Baudry J., Quere J. P., Fichet El., Relationships between landscape structure and the variations of common vole (Micmms arvalis) population, Landscape Ecol. 11 (1996) 279-288.

[25] den Boer P.J., On the survival of populations in a heterogeneous and variable environment. Oecologia 50 (1981) 39-53.

[26] Di Castri F., Younes T., Fonction de la diversitt biologique au sein de l’tcosysteme, Acta Oecologica 11 (1990) 429-444.

[27] Eastman J.R., IDRISI for windows. User’s guide, IDRISI Pro- duction, Clark University, 1995.

[28] Flather C. H., Brady S. J., Inkley D. B., Regional habitat appraisals of wildlife communities: a landscape-level evaluation of a resource planning model using avian distribution data, Landscape Ecol. 7 (1992) 137-147.

1291 Fumess R.W., Greenwood, J.J.D., Birds as monitors of environmental change, Chapman & Hall, London, 1993.

[30] Goldberg D.E., Miller T.E., Effects of different resource addi- tions on species diversity in an annual plant communit,y, Ecology 71 (1990) 213-225.

[3 I] Gounot M., Methodes d’etude quantitative de la vegetation, Ed Masson et Cie, Paris, 1969, 3 14 p.

[32] Gower J. C., Introduction to ordination techniques, in: Leg endre P., Legendre L. (eds), Developments in Numerical Ecology, NATO 14 (1987) 3-64.

1331 Gras R., Benoit M., Deffontaines J. P., Duru M., Lafarge M., Langlet A., Osty P.L., Le fait technique en agronomie. Activite agricole, concepts et methodes d’etude, INRA, L’Harmattan, 184 p.

1341 Huston M.A.. Biological diversity, Cambridge University Press, 1994,681 p.

[35] Klir G. J., Folger T. A., Fuzzy sets, uncertainty, and information, Prentice Hall, PTR, 1988, 355 p.

[36] Kozakiewicz M., Szacki J., Movements of small mammals in a landscape: patch restriction or nomadism? in: Lidicker W.2:. (ed.), Landscape approaches in mammalian ecology and conservation, University of Minnesota Press, 1995, pp. 78-94.

[37] Lande R., Statistics and partitioning of species diversity, and similarity among multiple communities, Oikos 76 (1996) 5- 13.

[38] Lebeaux M. O., ADDAD. Association pour le Developpement et la Diffusion de I’Analyse des Don&es, 1985, 195 p.

[39] Le Grand I., Dynamique du paysage de polders en Baie du Mont saint Michel. Universite de Paris IV, 1995, 2 15 p.

[40] MacArthur R. H., Environmental factors affecting bird species diversity. Am. Nat. 98 (1964) 387-398.

[41] Magurran A.E., Ecological diversity and its measurements, Princeton University Press, 1988, 215 p.

[42] McLaughlin A., Mineau P., The impact of agricultural practices on biodiversity, Agr. Eco. Env. 55 (1995) 201-212.

[43] McNeely J. A., Gadgil M., Ltveque C., Padoch C., Redford K., Human influences on biodiversity, in: Heywood V.H., Watson R.T. (eds), Global biodiversity assessment, Cambridge Univer- sity press, 1995, pp. 71 l-822.

[44] Medley K. E.. Okey W. O., Barrett G. W., Lucas M. E, Renwick W. H., Landscape change with agricultural intensification in .s rural watershed, southwestern Ohio, USA, Landscape Ecology 10 (1995) 161-176.

[45] Meeus J.H.A., Pan-European landscapes, Landscape and Urban Planning 3 I (1995) 57-79.

[46] Meeus J., Wijermans M., Vroom M., Agricultural landscapes in Europe and their transformation, Landscape and Urban Plan- ning 18 (1990) 289-352.

[47] Merriam G., Wegner J., Local extinctions, habitat fragmentation, and ecotones, Ecol. Stud. 92 (1992) 150-169.

[48] Michelat D., Giraudoux I?, Dimension du domaine vital de la chouette effraie Tyto alba pendant la nidification, Alauda 59 (1991) 137-142.

[49] Milne B. T., Heterogeneity as a multiscale characteristics of landscapes, in: J. Kolasa and S. T. A. Pickett (eds), Ecological heterogeneity, Springer-Verlag, New-York, 199 1, pp. 69-84.

[50] Morvan N., Structure et Biodiversitk de paysages de bocage : Le cas des empidides (Diptera, Empidoidea), Unpublished PhD thesis, Universite de Rennes 1, 1996.

[Sl] Morvan N., Delettm Y.R., Trehen P., Burel F., Baudry J., The distribution of Diptera Empididae in hedgerow network landscapes, British Crop Protection Conference Monographs, 58 (1994) 123-127.

[52] Norton B.G., The preservation of species. Princeton University Press, Princeton, 1986.

1531 Paillat G., Butet A., Spatial dynamics of the bank vole (Clethri- onomys glareohs) in a fragmented landscape, Acta Oecologica ( 1996).

[54] Paillat G., Butet A., Utilisation par les petits mammiferes du rtseau de digues bordant les cultures dans un paysage polderise d’agriculture intensive, Ecologia Mediterranea (in press).

[55] Paoletti M. G., Pimentel D., Biotic diversity in agroecosys- terns, Elsevier, 1992, 356 p.

[56] Probst J.R., Weinrich J., Relating Kirtland’s warbler population to changing landscape composition and structure, Landscape Ecol. 8 (1993) 257-273.

[57] Rozenberg R., Benthic fauna1 dynamics during succession following pollution abatement in a Swedish estuary, Oikos 27 (1976) 385-387.

[58] Rosenzweig M.L., Species diversity in space and time, Cam- bridge University Press, 1995, 436 p.

[59] Solbrig O.T., From genes to ecosystems : a research agenda for biodiversity, IUBS-SCOPE-UNESCO, 199 1, 124 p.

[60] Squires V.R. Dietary overlap between sheep, cattle and goats when grazing in common, J. Range. Manage. 35 (1982) 116-I 19.

[61] Taylor I., Predator-prey relationships and conservation, Cambridge University Press, Cambridge, 1994.

[62] Tew T.E., MacDonald D.W., Rands M.R.W., Herbicide applica- tion affects microhabitat use by arable wood mice (Apodemus sylvaticus), J. Appl. Ecol. 29 (1992) 532-539.

[63] Thenail C., Exploitations agricoles et territoire(s): contribu- tion 2 la structuration de la mosarque paysagke, Unpublished PhD Thesis, Universitd de Rennes 1, 1996.

[64] Turner M., Landscape heterogeneity and disturbance, Ecological studies 64, Springer Verlag New York, 1987, 239 p.

[65] Turner M. G., Gardner R.H., Quantitative methods in landscape ecology, Springer Verlag, 199 1, 536 p.

[66] Turner II, B. L., Meyer, W., Global land-use and land-cover change: an overview, in: Meyer W. B., Turner 11 B. L. (eds), Changes in land use and land cover: a global perspective, Cambridge University Press, Cambridge, 1994, pp. 3-10.

[67] Turner, S. J., O’Neill, R. V., Conley, W., Conley, M. R. & Humphries, H. C., Pattern and scale : statistics for landscape ecology, in: Turner M. G., Gardner R. H. (eds), Quantitative methods in landscape ecology, Springer-Verlag, 199 I, pp. 17-50.

Vol. 19(l) 1998

58 F. Burel et al.

[68] Van Apeldoom R.C., Oostenbrink W.T., Van Winden A., Van der Zee F.F., Effects of habitat fragmentation on the bank vole, Clethrionomys &reolu.s, in an agricultural landscape, Oikos. 65 (I 992) 265-274.

[69] Wiens J.A., Population responses to patchy environments. Ann. rev. Ecol. System. 7 (1976) 81-120.

[70] Wiens J. A.. Stenseth N. C., Van Horne B., Ims R. A., Ecological mechanisms and landscape ecology, Oikos 66 ( 1993) 369-380.

Acta Oecoiogicu


APPENDIX 1. Characteristics of types of windows (values are mean number of pixels per window)

A.l.l. Characteristics of 100 x 100 m windows

Class hedgerows woodlots grassland cropland hedgerow/ grassland

hedgerow/ cropland

I 39 6 305 I 55 3

2 61 2 212 91 61 26 3 14 128 85 93 33 26 4 1 0 33 346 0 2

5 59 6 65 243 22 55 6 31 3 16 328 3 44

A.1.2. Characteristics of 500 x 500 m windows


hedgerow/ cropland

1 61 0 933 8285 5 92

2 461 175 1080 7562 140 485

3 382 48 5163 3256 401 202

4 1419 422 4092 3550 I268 730

5 1515 366 2652 495 1 982 1142

6 1045 220 2109 5900 533 850

A.13 Characteristics of 750 x 750 m windows


hedgerow/ cropland

1 130 0 2055 18702 9 200

2 1047 288 3169 16181 386 1063

3 1818 718 10218 7911 1615 1081

4 2896 698 5977 11492 I882 2083

Vol. 19 (1) 1998

60 F. Burel et al.

APPENDIX 2. Relationships between types of windows and landscape units.

A.2.1. Classes of 100 x 100 m windows Redundancy = 23.2 %

Class Unit A Unit B Unit C Unit D

I IX.1 9.4 10.9 I .8

2 16.9 14.5 5.8 0

3 IO.5 6.6 2.4 0

4 5.6 16.7 49.2 x9

5 30.3 28.9 10.7 I .7

6 IX.4 23.8 20.‘) 7.4

A.2.2. Classes of 500 X 500 m windows. Redundancy = 58.4 %


I 0 0 2.1 97

2 0 11.4 71.1 2.4

3 2 0 12.4 0.7

4 22 IO.5 0 0

5 41 24.6 0 0

6 17 53.5 14.4 0

A.2.3. Classes of 750 x 750 m windows. Redundancy = 76.8 %


I 0 0 0 95.3

2 0 13.8 91.5 4.7

3 7.9 0 8.5 0

4 92.1 86.2 0 0

Comparative biodiversity along a gradient of agricultural landscapes

Documents